Evaluation of the Jokoo-ION 150 AC: Guidelines for the evaluation of analysers by ion-selective electrodes

The Jokoo-ION 150 AC is an automatic sodium, potassium and chloride analyser which uses ion-selective electrodes. The sample mode can be whole-blood, serum or urine. To evaluate a urine sample, a previous dilution (1:6) with the standard 1 solution is required. For three concentrations of control materials, the total precision (CV) ranged from 0.17 to 1.22% for sodium, 0.22 to 1.69% for potassium and 0.16 to 0.74% for chloride. The system demonstrated acceptable performance in detection limit, linearity, drift and carry-over. Patients' results from ION 150 AC correlated well with those from a SMAC II and an IL 943. In a study on potential interferences, a slightly negative interference with potassium was found with increases in lithium; and only high ammonium ion concentrations caused a positive interference on the potassium ion. A slightly positive interference of bromide on the chloride ion was also found. The electrode slope, electrode response, sample temperature and pH effect, effects of high concentrations of proteins or lipids, and haematocrit influence on the sodium, potassium and chloride ion concentration were also evaluated. The strategy adopted in this study provides an ideal framework for future evaluations of ionselective electrode analysers.

The system demonstrated acceptable performance in detection limit, linearity, drift and carry-over. Patients' results from ION 150 A C correlated well with those from a SMA C H and an IL 943. In a study on potential interferences, a slightly negative interference with potassium was found with increases in lithium; and only high ammonium ion concentrations caused a positive interference on the potassium ion. A slightly positive interference of bromide on the chloride ion was also found.
The electrode slope, electrode response, sample temperature and pH effect, effects of high concentrations of proteins or lipids, and haematocrit influence on the sodium, potassium and chloride ion concentration were also evaluated. The strategy adopted in this study provides an ideal framework for future evaluations of ionselective electrode analysers. the selective membrane by an electrolytically simple solution of the ion to be evaluated. The difference of potential measured at the extreme points of the system is derived from the equation of Nernst: RT E E0 + Ep + 2"3 --logA [1] E0 is the sum of the fixed potentials depending on the measuring and reference electrodes structure, E/ is the sum of the variable potentials, 'parasites', that have to be annulled, minimized or at least stabilized, R is the perfect gases constant, Tis absolute temperature, F is Faraday's number, and n is valence. A=yxC logy 0"509 n 2 X/I I 1/2 Z(Cini) [2] where Ci is concentration, ni is valence (of each ion present in the solution),A is the activity, y is the activity coefficient, C is the molar concentration, and I is the ionic force.
At very dilute concentrations (10 -7 M), the activity coefficient is very near to unity but at higher concentrations it diminishes [3 and 14]. Thus, the calibrator must have an activity coefficient similar to that of the sample.

Introduction
The device which is commonly called an 'ion-selective electrode' ('ISE') determines the activity (A) of an ion in solution (in clinical biochemistry it usually refers to cations). The essential part of this device is a membrane with properties related to the ion being evaluated: it is specifically, or at least selectively, permeable to this ion. Therefore, a potential difference is established between both sides of the.membrane, which is proportional to the ion activity difference. The technology behind electrometric instruments aims to make the difference in potential as small as possible.
As a result, a measuring chain is used where a series of different devices are capable of exploiting this small potential. These complementary devices are called reference electrodes, and are characterized by the stability of their own potential. There are two classes of reference electrode; silver chloride and mercury chloride (calomel).
The chain of measurement is composed of a sequence including a reference electrode (called 'external') and a measuring electrode-the latter being a reference electrode designated 'internal', which is separated from Ion-selective electrodes have a big international market due to their simplicity in use and their inclusion in automatic analysers.
The unequal quality of the sensors currently available, together with the sparse experience that analysis laboratories have in the field of electrometry and the limited quality control materials at present available from manufacturers, means that these instruments are often difficult to work with.
The present paper presents an evaluation of an ionselective electrode analyser (the Jokoo-ION 150 AC) and suggests guidelines for studying, controlling and evaluating ion-selective electrode analysers.

Materials and methods
Automatic analyser SMA C H (Technicon Corporation, Tarrytown, New York 10591, USA) Instruments SMAC II is a continuous flow analyser, which evaluates sodium and potassium ions by indirect potentiometry. For sodium the total dilution is 16"2, and for potassium 1:18"8.
The sodium ion utilizes a sodium-aluminium glass electrode where sodium selectivity against potassium amounts to 1000" 1; the serum sample is buffered at pH 8"0 to eliminate any interference by the hydrogen ion [4].
The Valynomicine electrode used for the potassium ion has a selectivity of nearly 4000:1 against the sodium ion. Valynomicine is an antibiotic which is insoluble in water, but is soluble in organic solvents, such as ether or acetone. It is suspended at the end of the electrode in a porous membrane. The inner solution, formed by potassium chloride, is in contact with the Ag-C1Ag inner medium from the reference electrode [5].

Dugnano, Italy)
The IL 943 has a dilutor incorporated, providing automatic calibration and sampling. The calibrator and sample dilution is 1:100. It is made with a caesium chloride solution, which has final concentration of 1.5 mmol/1. The instrument's zero is regulated with distilled water at 100 with the inner standard (caesium chloride solution [7]. The gas used is 99% pure propane at 3"6 psi.

Jokoo-ION 150 A C analyser
This analyser measures sodium, potassium and chloride ion concentrations simultaneously for every sample through selective electrodes by direct potentiometry.
The sample volume needed is 100 1, plus 50 tl residual volume. This sample can be whole-blood, plasma or urine. To evaluate a urine sample, a previous dilution (1:6) with standard solution is required. The throughput is 150 samples/hour.
When the system is ready, a complete calibration is realized, being recalibrated automatically at two points every 30 min (this interval is optional).
The automatic sampler holds 30 sample cuvettes in a gyratory disk, allowing positional identification of the sample. This unit can use manual sampling through syringes or capillaries, and stat sampling. It has an automatic sample detector. Results are obtained in visual digits and are also printed-out. The sample is aspirated through a capillary and is then transported directly to the electrode unit by a peristaltic pump which is placed after the measuring unit. The capillary is rinsed automatically with standard solution 1.
There are four electrodes in the measuring unit: sodium, potassium, chloride and the reference electrode. The reference electrode system is a double union type (liquid-liquid) (Ag/AgC1 in saturated KC1 solution), which provides a stable potential that can be used as a reference. The sodium electrode is a glass membrane electrode (sodium aluminium silicate). The potassium electrode is a liquid membrane electrode, which is based on a neutral carrier (Valynomicine-PVC, polyvinyl chloride as a support). Finally, the chloride electrode is a liquid membrane electrode, which is based on ionic exchange using quaternary ammonium salts as an ion-exchanger in polymeric solvent.. Also used in the evaluation were a tester (Kaiser model SK-6300); a voltimeter of 1014 Ohms and a sensitivity of 0"1 milivolt (mV); and an analogous registrator (Linear Instruments Corporation Model 1201; scale mV-5 V).

Controls
Three commercial lyophilized controls ofhuman origin at three concentrations (low, medium and high) were used.
These were selected to represent common clinical decisions. The manufacturer's recommendations were followed to reconstitute the lyophilized control. A pool was prepared and stored in aliquot parts in the freezer (-20 C). As urine control, a pool of urine was used to study the within-run imprecision.

Reagents
(1) Kit for determination of ammonium in blood (SIGMA Diagnostics, St. Louis, USA). This method is based on the ammination of 2oxoglutarate by means of glutamate dehydrogenase (GLDH) and reduced nicotinamide adenine dinucleotide (NADH). The absorbance decrease, measured at 340 nm, as a result of the NADH oxidation, is proportional to the concentration of plasma.

Imprecision
Within-run imprecision A sample of each of the three controls was analysed 20 times in the same run without interruption for calibration or recovery of the base-line. This process was repeated for   125"14 0"53 two working days. The mean, standard deviation (SD) and the coefficient of variation (CV) on each concentration level were calculated.
The urine pool was prepared in the laboratory; it was diluted at 1" 6 in standard solution 1, and was evaluated 30 times within the same run/test. The mean, SD and CV were also calculated.
The results are shown in table 1" precision for the three ions for all three concentration levels was high as was precision with the urine pool.

Day-to-day imprecision
One sample of each of the three control concentration levels was evaluated each day over 20 days. The mean, SD and CV were computed for each of the three concentration levels. The results are presented in table 2. A high precision for the three components in the three studied concentration levels is shown.

Inaccuracy
Because there were no commercial controls available [br evaluating the system, primary standards at three concentration levels were prepared by weighing them into buffer Tris-orthophosphoric acid (Tris 0"2 M, orthophosphoric acid 0"15 N), adjusted to pH 7"35, and adding a surfactant (brij 30%) at ml/1. Table 3 shows the concentration levels which were examined. A duplicate determination was carried out and the theoretical value was checked by flame photometry. To calculate the inaccuracy percentage, the following formula was used:   This process was repeated for nine days, and the inaccuracy was calculated taking the mean of the measured values. Good accuracy was observed for the sodium and the potassium ions, although the system tended to give values which were inferior to the theoretical ones. The results obtained for chloride ions were substantially below the theoretical ones.
The correlation between direct potentiometry and flame photometry and the indirect potentiometry (SMAC II) was also studied. One hundred serum samples of patients with concentrations varying from the studied components were assessed, in duplicate and by the three methods. The duplicates were distributed throughout the series, including lipaemic, icteric and haemolitic samples. In addition, 90 urine samples were processed on the same day, by both the Jokoo analyser and by flame photometry.
The correlation coefficient was calculated, the least squares line and the least squares regression according to Deming [9]. The results are shown in the table 4. The detection limit was taken to mean the least quantity, or concentration, of a substance that can be distinguished with a given probability of a reaction blank carried out under the same conditions [10]. This ideal reaction blank is executed by means of the Tris-orthophosphoric acid buffer at pH 7"35 and without the ions to be determined.
The ionic concentrations of sodium, potassium and chloride were determined in a series of 10 determinations in 3 days (thus obtaining a total of 30 values per ion). For arisk oc to 5% and large samples (V> 30) the upper limit of the blank confidence interval was calculated as being 2 + "97, a value that corresponds to the detection limit for this risk. The results are shown in table 5.

Lineariy limits
Linearity is expressed by the equation y= a + bx [10].
Because it was impossible to obtain a standard containing the pure ion to be determined, sodium chloride and potassium chloride were used. On the basis of a stock standard prepared by weighing in Tris-orthophosphoric acid buffer at pH 7"35, a series in triplicate of decreasing dilutions was pertbrmed with the same buffer up to a total of 12 points, which were evaluated on three different days with a total of nine determinations per point. Then for each point the mean value and the standard deviation were calculated. The mean of the observed values corresponds to the ordinate axis, and the theoretical values to the abscissa axis. The bisector (y bx, b 1) is drawn and the linearity limits correspond to the maximum and minimum points ot' which the interval of confidence cut the bisector. The interval was calculated for a risk oc of ,5% and 8 degrees of freedom by X +_ 2.306 SD.
The least squares line and the correlation coefficient corresponding to the linearity interval was obtained. The results are shown in table 6.
Estimation qf the total error The errors corresponding to the inaccuracy and imprecision of a technique, which can sum up or compensate each other, can be calculated as follows: A-I x SD and A + x SD (10) where A inaccuracy, SD standard deviation of the day-to-day imprecision, and student value according to the risk oc and the degrees of freedom. Table 7 shows the results.

Carry-over
The evaluations of carry-over used three protocols [1 1-13] -each author reports on different aspects, all of which can affect the contamination between samples evaluated in series. For clarity, this report describes each of the three protocols, and results are given for each method.
Primary standards were prepared in the same way as those used in the linearity and inaccuracy study.      The following results were obtained: K 0"038% for sodium; K 0% for potassium; K 1"47% for chloride. Contamination, then, is practically negligible for the three studied components.

Drift
A commercial control was usedthe level of which coincided with the physiological values for each component. Controls were evaluated at the beginning of a series of samples (after the first calibration), at the end of this series of samples (prior to the recovery of the base-line and to the second calibration) and, finally, at the end of the day's work after having processed various sample Log. concentration Therefore, the drift has been studied both in a series and also during the day's work. The values obtained at the beginning of the series were compared to those at the end of the series. In addition, those values obtained at the beginning of the series were compared to those from the end of the day's work.  [1][2][3]. The abscissa axis is the concentration logarithm and the ordinate axis the corresponding mVvalue; the slope of every electrode was obtained by giving it the mV-values corresponding to 10-times increase in the concentration of the ion (a 'decade') [14].  1000 Electrode response study The response of each electrode was studied separately, using a serum pool and concentrating on the response time and the stability of the steady state obtained [15].
For that purpose an analogue registrator was connected to the outlet of every electrode. The stability of the steady state was good for all three electrodes, with a very short response time-less than s-in achieving the plateaus and with a total duration of less than 22 s.
The results are shown in figures 4, 5 and 6.

Sample temperature effect
Three commercial controls were evaluated on being removed from the refrigerator (-4 C), and after being stored at room temperature (26 C). The controls were all evaluated after 5, 15 and 30 min, and always within the same analytic series.
The percentage of variation for each concentration level was calculated, with reference to the concentration which was evaluated initially (--4C). In a similar way the pH levels of interest to clinical practice were measured, and a minimm influence within lhe pH-margins were observed -see table 11. Interference studies Influence of the ,;odium ion on the potassium ion electrode. A series of solutions were prepared with a constant concentration of 4 retool/1 of potassium and decreasing sodium concentrations of 200, 100, 50 and 8 mmol/1, in Tris-orthophosphoric acid buffer with ml/1 of a surfactant. Each solution was evaluated five times, the mean value taken and the effect of the sodium ion concentration over the constant value of the potassium ion calculated. The concentrations of this series were also evaluated by flame photometry, in order to check the accuracy of the concentration of each ion.
In figure 10  values against the variations of the sodium ion concentration, can be observed. Between the margins corresponding to 50 mrnol/1 and 200 mmol/1 of sodium, the effect on the potassium ion has no significance.
The same sample series was processed by the ion analyser, tbllowing the protocol fbr urine samples, and no effect was found on the series.
Effect of the polassium ion on the sodium ion electrode. For a constant concentration of 125 mmol/1 sodium, potassium concentrations of 250, 125, 62, 30, 15, 7"5 and 3"2 retool/1 in Tris-orthophosphoric acid buffer were prepared. Each solution was evaluated five times, and the mean value was determined later.
The potassium ion concentration effect on a constant sodium ion concentration was also calculated. The accuracy of the concentrations of both ions in this series was checked by flame photometry. Figure 11 shows the potassium ion values on its abscissa and the sodium ion values, found as a result of the different potassium ion concentrations, on ts ordinate axis. The influence of the potassium ion on the electrode of the sodium ion is insignificant.
Influence oj the lithium ion as   Figure 13 gives the sodium and potassium concentrations on its ordinate axis, against ammonium ion concentrations. Neither ion is subject to any interference from ammonium ions at physiopathological concentrations. Significant interference was seen with experimental concentrations of ammonium higher than 10 mmol/1 (1"38%) and 100 mmol/1 (28.7%), which might affect the mV is a geometrically proportional influence of bromide on chloride, which, in spite of the fact it is small at lower concentrations, should be taken into account for patients who are receiving bromide and other halogens in substantial amounts.  A human albumin solution at 20% was used which was poor in ions. A lipaemic solution which was 10% free of ions was also used. Decreasing dilutions were prepared, in duplicate, maintaining a fixed ion concentration and processing each of them following three survey methods. Tables 12 and 13 show the variation of direct potentiometry in relation to indirect potentiometry and flame photometry.

Influence of lipid
Due to the effect of the sample content of plasmatic water [16], some sodium and potassium ion values could be observed to be inversely proportional to the protein concentration in those methods that employ dilution relating to direct potentiometry. The same effect, but to a minor degree, could be seen in the 10% lipid series (longchain triglyceride).
Whole blood samples Haematocrit influence. The eventual haematocrit influence on the sodium and potassium and chloride ion concentration evaluation were checked. Starting from blood samples obtained with lithium heparin anticoagulant of the same patient, a series of whole blood samples was prepared with a haematocrit varying from 4 to 70%. This preparation was made by adding or subtracting plasma, after the corresponding phases of centrifugation, until the desired haematocrits were achieved.
In this assay, very small differences in the chloride ion, potassium ion and sodium ion evaluations were observed, except for the samples with a haematocrit higher than 65%, where the potassium ion concentration value was subject to a strong increase. This effect could result from the way the sample flows through the capillary which  (see table  14).

Conclusions
The evaluation described here includes guidelines for ionselective electrode analyser evaluations. The following results were found from the assays reported: Sodium (mmol/L) Figure 10. Influence of the sodium ion on the potassium electrode.
(2) The inaccuracy relating to the primary standard values was good for the sodium and potassium ion, and acceptable tbr the chloride ion.
(4) The linearity limits are tfigher than 500 mmol/1 for the sodium ion and chloride ion, and more than 45 mmol/1 for the potassium ion.
(5) Carry-over and drift are negligible (smaller than 2%). The results were not significantly affected by sample temperature. (6) In. the study of the electrode slope, some values close to 59 mV were obtained, which is considered to be the optimal value for monovalent ions. The response time is also very short (less than s), and stability is Inaintained for 22 s.
(7) Significant variations due to.pH were observed on both ions (sodium and potassium). (8) The interference study showed a slightly negative interference on the potassium ion due to the increment of the lithium ion. Only high ammonium ion concentrations cause a positive interference on the potassium ion. It was also observed that there    (10) In the evaluation of whole blood samples significant increases of the potassium values were observed, starting from a haematocrit of 60% on.